US-20260127648-A1 - USING TRAINED MACHINE-LEARNING MODEL OF AN ONLINE SYSTEM TO PREDICT TIMING OF STATE CHANGE OF VARIABLE STATE ITEM

Abstract

An online system uses a trained machine-learning model to predict timing of a state change of a variable state item in an order. The online system applies a trained machine-learning model to information about the variable state item and information about an ambient condition when servicing the order to predict a timing when a state of the variable state item changes from an original state at a location of a source associated with the online system to a different state. Based on the predicted timing, the online system generates a control signal that initiates at least one of a first action associated with the order or a second action associated with the variable state item. The online system performs, using the control signal, at least one of the first action associated with the order or the second action associated with the variable state item.

Inventors

• Brent Scheibelhut
• Sara Starck
• Clyde Simmons Manuel
• Brandon Sim
• Karen Kraemer LOWE
• Erica Jazayeri Quintana
• Justin Kuo-Ting Tsung
• Richard Lam

Assignees

• MAPLEBEAR INC.

Dates

Publication Date

20260507

Application Date

20241107

Claims (20)

1. 1 . A method, performed at a computer system comprising a processor and a computer-readable medium, comprising: receiving, via a network from a device associated with a user of an online system, a signal that triggers a process of servicing an order; identifying an item from the order having a state that is varying over time; accessing a state change prediction machine-learning model of the online system, wherein the state change prediction machine-learning model is trained to predict a timing when a state of the item changes from an original state at a location of a source associated with the online system to a different state; applying the state change prediction machine-learning model to information about an ambient condition when servicing the order and information about the item to generate the timing when the state of the item changes from the original state to the different state; generating, based at least in part on the timing, a control signal that initiates at least one of a first action associated with the order or a second action associated with the item; and performing, using the control signal, at least one of the first action associated with the order or the second action associated with the item.
2. 2 . The method of claim 1 , further comprising: receiving, from a device of an agent of the online system servicing the order and via the network, the information about the ambient condition including an ambient temperature; and retrieving, from a database of the online system, the information about the item including a set of temperature features for the item.
3. 3 . The method of claim 1 , further comprising: receiving, from a device of an agent of the online system servicing the order and via the network, information about the agent including one or more images of a cooling device in a vehicle of the agent that will be used for servicing the order, wherein applying the state change prediction machine-learning model comprises applying the state change prediction machine-learning model further to the information about the agent to generate the timing when the state of the item changes from the original state to the different state.
4. 4 . The method of claim 1 , further comprising: receiving, from the device associated with the user and via the network, information about the user including information about a current location of the user, wherein applying the state change prediction machine-learning model comprises applying the state change prediction machine-learning model further to the information about the user to generate the timing when the state of the item changes from the original state to the different state.
5. 5 . The method of claim 1 , wherein the control signal includes a user interface signal, and the method further comprising: sending the user interface signal to at least one of a device associated with an agent of the online system servicing the order or the device associated with the user, wherein the sending causes at least one of the device associated with the agent to display a first user interface or the device associated with the user to display a second user interface with an indication of at least one of the first action or the second action.
6. 6 . The method of claim 1 , wherein performing the first action comprises: determining, based at least in part on the timing, a schedule for servicing the order; and sending, via the network, the control signal including a user interface signal to a device associated with an agent of the online system servicing the order, wherein the sending causes the device associated with the agent to display a user interface with an indication of the first action including information about the schedule for servicing the order.
7. 7 . The method of claim 1 , wherein performing the first action comprises: splitting, based at least in part on the timing, servicing of the order into servicing a first part of the order including the item during a first time period and servicing a second part of the order without the item during a second time period that is later than the first time period; and sending, via the network, the control signal including a user interface signal to a device associated with an agent of the online system servicing the order, wherein the sending causes the device associated with the agent to display a user interface with an indication of the first action including information about servicing the first part of the order during the first time period and servicing the second part of the order during the second time period.
8. 8 . The method of claim 1 , wherein performing the second action comprises: cancelling, based at least in part on the timing, the item from the order; generating, using the control signal, an appeasement for the item; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display a user interface with an indication of the second action including information about cancelling the item from the order and a notification about the appeasement for the item.
9. 9 . The method of claim 1 , wherein performing the first action comprises: eliminating, from a set of time periods for servicing the order and based at least in part on the timing, one or more time periods; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display a user interface with an indication of the first action including information about the set of time periods for servicing the order without the one or more time periods.
10. 10 . The method of claim 1 , further comprising: determining, based at least in part on the timing and the information about the ambient condition when servicing the order, a first placement in a user interface of the device associated with the user for a first set of items having stable states over time and a second placement in the user interface for a second set of items having varying states over time, the second placement lower than the first placement; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display the user interface with the first set of items at the first placement and the second set of items at the second placement.
11. 11 . The method of claim 1 , wherein performing the second action comprises: accessing an item removal machine-learning model of the online system, wherein the item removal machine-learning model is trained to identify whether the item should be removed from the order; applying the item removal machine-learning model to at least one of first feedback information about the item provided by a plurality of users of the online system, one or more features of an agent of the online system servicing the order, the information about the ambient condition, or second feedback information about the item provided by the user to generate a removal score for the item that is indicative of whether the item should be removed from the order; cancelling, based on the removal score, the item from the order; generating, using the control signal, an appeasement for the item; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display a user interface with an indication of the second action including information about cancelling the item from the order and a notification about the appeasement for the item.
12. 12 . The method of claim 1 , further comprising: generating training data including labels collected in lab settings about changes of states of a plurality of items over time at a plurality of ambient temperatures; and training, using the training data, the state change prediction machine-learning model to generate a set of initial values for a set of parameters of the state change prediction machine-learning model.
13. 13 . The method of claim 1 , further comprising: receiving, from the device associated with the user and via the network, feedback data with information about a score provided by the user via a user interface of the device associated with the user about the state of the item at a location of the user; and re-training the state change prediction machine-learning model by updating, using the feedback data, a set of parameters of the state change prediction machine-learning model.
14. 14 . A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor, cause the processor to perform steps comprising: receiving, via a network from a device associated with a user of an online system, a signal that triggers a process of servicing an order; identifying an item from the order having a state that is varying over time; accessing a state change prediction machine-learning model of the online system, wherein the state change prediction machine-learning model is trained to predict a timing when a state of the item changes from an original state at a location of a source associated with the online system to a different state; applying the state change prediction machine-learning model to information about an ambient condition when servicing the order and information about the item to generate the timing when the state of the item changes from the original state to the different state; generating, based at least in part on the timing, a control signal that initiates at least one of a first action associated with the order or a second action associated with the item; and performing, using the control signal, at least one of the first action associated with the order or the second action associated with the item.
15. 15 . The computer program product of claim 14 , wherein the instructions further cause the processor to perform steps comprising: receiving, from a device of an agent of the online system servicing the order and via the network, the information about the ambient condition including an ambient temperature; and retrieving, from a database of the online system, the information about the item including a set of temperature features for the item.
16. 16 . The computer program product of claim 14 , wherein the instructions further cause the processor to perform steps comprising: determining, based at least in part on the timing, a schedule for servicing the order; and sending, via the network, the control signal including a user interface signal to a device associated with an agent of the online system servicing the order, wherein the sending causes the device associated with the agent to display a user interface with an indication of the first action including information about the schedule for servicing the order.
17. 17 . The computer program product of claim 14 , wherein the instructions further cause the processor to perform steps comprising: splitting, based at least in part on the timing, servicing of the order into servicing a first part of the order including the item during a first time period and servicing a second part of the order without the item during a second time period that is later than the first time period; and sending, via the network, the control signal including a user interface signal to a device associated with an agent of the online system servicing the order, wherein the sending causes the device associated with the agent to display a user interface with an indication of the first action including information about servicing the first part of the order during the first time period and servicing the second part of the order during the second time period.
18. 18 . The computer program product of claim 14 , wherein the instructions further cause the processor to perform steps comprising: determining, based at least in part on the timing and the information about the ambient condition when servicing the order, a first placement in a user interface of the device associated with the user for a first set of items having stable states over time and a second placement in the user interface for a second set of items having varying states over time, the second placement lower than the first placement; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display the user interface with the first set of items at the first placement and the second set of items at the second placement.
19. 19 . The computer program product of claim 14 , wherein the instructions further cause the processor to perform steps comprising: accessing an item removal machine-learning model of the online system, wherein the item removal machine-learning model is trained to identify whether the item should be removed from the order; applying the item removal machine-learning model to at least one of first feedback information about the item provided by a plurality of users of the online system, one or more features of an agent of the online system servicing the order, the information about the ambient condition, or second feedback information about the item provided by the user to generate a removal score for the item that is indicative of whether the item should be removed from the order; cancelling, based on the removal score, the item from the order; generating, using the control signal, an appeasement for the item; and sending, via the network, the control signal including a user interface signal to the device associated with the user, wherein the sending causes the device associated with the user to display a user interface with an indication of the second action including information about cancelling the item from the order and a notification about the appeasement for the item.
20. 20 . A computer system comprising: a processor; and a non-transitory computer-readable storage medium having instructions that, when executed by the processor, cause the computer system to perform steps comprising: receiving, via a network from a device associated with a user of an online system, a signal that triggers a process of servicing an order; identifying an item from the order having a state that is varying over time; accessing a state change prediction machine-learning model of the online system, wherein the state change prediction machine-learning model is trained to predict a timing when a state of the item changes from an original state at a location of a source associated with the online system to a different state; applying the state change prediction machine-learning model to information about an ambient condition when servicing the order and information about the item to generate the timing when the state of the item changes from the original state to the different state; generating, based at least in part on the timing, a control signal that initiates at least one of a first action associated with the order or a second action associated with the item; and performing, using the control signal, at least one of the first action associated with the order or the second action associated with the item.

Description

BACKGROUND Online systems often face a vast amount of user complaints, appeasements, and poor experiences due to non-shelf stable items (i.e., perishable items) arriving in a poor state (e.g., melted or otherwise spoiled) due to environmental or fulfillment factors. Therefore, it is desirable to reduce the number of appeasements due to perishable items being delivered in an unacceptable state (e.g., melted ice cream). However, there is a technical problem of how to reduce, in an automatic manner and at a large scale as required by an online system, a number of occurrences of perishable items being delivered in unacceptable states. In particular, strategies for reducing occurrences of perishable items being delivered may involve first predicting the timing of the perishable items changing from an acceptable state to an unacceptable one. However, there are no satisfactory techniques for predicting this timing for different types of items under varying delivery times and conditions. SUMMARY Embodiments of the present disclosure are directed to using a trained machine-learning model of an online system to predict timing of a state change of a variable state item (e.g., perishable item). Based on the predicted timing of the state change of the variable state item, the online system may generate a user interface that displays servicing modifications for an online order caused by the variable state item in the online order. The servicing modifications may include an increased ranking of stable state items in the user interface, elimination of delivery slots for the order with the variable state item, and/or removing and refunding the variable state item from the order. Additionally, based on the predicted timing of the state change of the variable state item, the online system may apply some other remedial action, such as making a different batching decision for the order with the variable state item, and/or modifying the order by splitting delivery of the order into separate delivery of stable state items and the variable state item. In accordance with one or more aspects of the disclosure, the online system receives, via a network from a device associated with a user of the online system, a signal that triggers a process of servicing an order. The online system identifies an item from the order having a state that is varying over time. The online system accesses a state change prediction machine-learning model of the online system, wherein the state change prediction machine-learning model is trained to predict a timing when a state of the item changes from an original state at a location of a source associated with the online system to a different state. The online system applies the state change prediction machine-learning model to information about an ambient condition when servicing the order and information about the item to generate the timing when the state of the item changes from the original state to the different state. The online system generates, based at least in part on the timing, a control signal that initiates at least one of a first action associated with the order or a second action associated with the item. The online system performs, using the control signal, at least one of the first action associated with the order or the second action associated with the item. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 illustrates an example system environment for an online system, in accordance with one or more embodiments. FIG. 2 illustrates an example system architecture for an online system, in accordance with one or more embodiments. FIG. 3 illustrates an example architectural flow diagram of using a trained machine-learning model of an online system to predict timing of a state change of a perishable item in an order, in accordance with one or more embodiments. FIG. 4 illustrates an example architectural flow diagram of using a trained machine-learning model of an online system to determine whether to remove an item from an order that can be detrimental to a state of the order when the order is delivered, in accordance with one or more embodiments. FIG. 5 is a flowchart for a method of using a trained machine-learning model of an online system to predict timing of a state change of a perishable item in an order, in accordance with one or more embodiments. DETAILED DESCRIPTION FIG. 1 illustrates an example system environment for an online system 140, in accordance with one or more embodiments. The system environment illustrated in FIG. 1 includes a user client device 100, a picker client device 110, a source computing system 120, a network 130, and an online system 140. Alternative embodiments may include more, fewer, or different components from those illustrated in FIG. 1, and the functionality of each component may be divided between the components differently from the description below. Additionally, each component may perform their respective functionalities in response to a request from a human,